Research map / 3D vision / 3D generation
JointHOI: Jointly Generating Contact Maps Enhances Hand Object Interaction Generation
Mingyeong Song, Jungbin Cho, Jisoo Kim, Ananya Bal, Kartik Sharma, Youngjae Yu, Laszlo Jeni, Junhyug Noh
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace two-stage pipeline with one-stage joint generation (ablation) | JointHOI baseline · ARCTIC | Top-1 accuracy: 0.520 -> 0.795 | Improved |
| add unified distance-based dynamic contact map (ablation) | JointHOI one-stage · ARCTIC | FID: 0.274 -> 0.145 | Improved |
| split dynamic contact map per hand (L/R) (ablation) | JointHOI with unified contact map · ARCTIC | FID: 0.145 -> 0.038 | Improved |
| add Contact Inner Guidance at inference (ablation) | JointHOI · ARCTIC | Interpenetration volume (cm3): 6.988 -> 4.406 | Improved |
| add Contact Inner Guidance at inference (ablation) | JointHOI · ARCTIC | Top-1 accuracy: 0.935 -> 0.948 | Improved |
| add Contact Inner Guidance at inference (ablation) | JointHOI · ARCTIC | Contact ratio (%): 92.21 -> 93.71 | Improved |
| add unified contact map to one-stage model (ablation) | JointHOI one-stage · ARCTIC | Interpenetration volume (cm3): 7.057 -> 7.146 | No clear effect |
| one-stage joint generation vs two-stage (ablation) | JointHOI baseline · ARCTIC | FID: 0.291 -> 0.274 | No clear effect |
| JointHOI vs Text2HOI | JointHOI · GRAB | Top-1 accuracy: 0.711 (Text2HOI) vs 0.663 | Hurt |
| JointHOI vs Text2HOI | JointHOI · GRAB | Interpenetration volume (cm3): 7.790 -> 3.419 | Improved |
| JointHOI vs Text2HOI | JointHOI · ARCTIC | FID: 0.148 -> 0.033 | Improved |
| use Log-l1 contact consistency loss in CIG instead of other losses (ablation) | CIG · ARCTIC and GRAB | Top-1 accuracy and ID | Improved |
Closest papers by meaning
- Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction (CVPR 2024 · similarity 0.94)
- How Do I Do That? Synthesizing 3D Hand Motion and Contacts for Everyday Interactions (CVPR 2025 · similarity 0.91)
- HandDiffuse: Generative Controllers for Two-Hand Interactions via Diffusion Models (AAAI 2025 · similarity 0.90)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.90)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.89)
- MEgoHand: Multimodal Egocentric Hand-Object Interaction Motion Generation (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
